0
$\begingroup$

I wonder if my following understandings about random seeds are correct:

  1. Uniform random generators require random seeds.

  2. Non-uniform random generators depend on random seeds, only through uniform random generators.

  3. Or are there (common) non-uniform random generators use random seeds directly, without indirectly via uniform random generators?

Thanks!

$\endgroup$
  • $\begingroup$ Uniform random number generators require a seed for initialization, not necessarily a random seed. In fact, at times, it is useful for debugging purposes to be able to specify that the same seed be used so that the program being debugged does exactly the same calculations as it did the previous time (except in places where the programmer has made changes on the program between the two runs). In many instances, if a seed is not specified (as in an initialization call to rand() instead of to rand(1234567890)), the system clock is read and used as the seed on the first invocation of rand. $\endgroup$ – Dilip Sarwate Sep 24 '12 at 12:21
  • $\begingroup$ Just to add to dilip's comment which is actually a good answer. Nonuniform random numbers can always be obtained from uniform random numbers. This can usually be done using the inverse cumulative distribution transformation. $\endgroup$ – Michael Chernick Sep 24 '12 at 12:38
  • $\begingroup$ @DilipSarwate: From the Wikipedia link, a random seed needs not be random, or do I miss something? $\endgroup$ – Tim Sep 24 '12 at 12:49
  • $\begingroup$ @MichaelChernick: I wonder if a nonuniform random number generator also can rely on random seed but not via uniform random number generator? $\endgroup$ – Tim Sep 24 '12 at 12:51
  • $\begingroup$ whatever way you generate pseudorandom numbers if the method is iterative and so is a function of the previous random number (hence the term pseudo it will require a seed to get started. Random seeds are only used when there is a desire to generate sets of random numbers that don't go through the same sequence. For replicating results or at least the random numbers in your simulation you would use a fixed seed so that the sequence will be repeatable. $\endgroup$ – Michael Chernick Sep 24 '12 at 13:19
2
$\begingroup$

From the comment by @Dilip Sarwate

Uniform random number generators require a seed for initialization, not necessarily a random seed. In fact, at times, it is useful for debugging purposes to be able to specify that the same seed be used so that the program being debugged does exactly the same calculations as it did the previous time (except in places where the programmer has made changes on the program between the two runs). In many instances, if a seed is not specified (as in an initialization call to rand() instead of to rand(1234567890)), the system clock is read and used as the seed on the first invocation of rand.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.